A Phylogenetic and Biogeographic Study of the ( Tribe Oenantheae) Author(s) :Tiffany S. Bone, Stephen R. Downie, James M. Affolter, and Krzysztof Spalik Source: Systematic Botany, 36(3):789-805. 2011. Published By: The American Society of Taxonomists URL: http://www.bioone.org/doi/full/10.1600/036364411X583745

BioOne (www.bioone.org) is a a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research. Systematic Botany (2011), 36(3): pp. 789–805 © Copyright 2011 by the American Society of Plant Taxonomists DOI 10.1600/036364411X583745 A Phylogenetic and Biogeographic Study of the Genus Lilaeopsis (Apiaceae tribe Oenantheae)

Tiffany S. Bone ,1,4 Stephen R. Downie ,1, 5 James M. Affolter ,2 and Krzysztof Spalik3 1 Department of Plant Biology, University of Illinois at Urbana-Champaign, Illinois 61801 U. S. A. 2 College of Agricultural and Environmental Sciences, University of Georgia, Athens, Georgia 30605 U.S.A. 3 Department of Plant Systematics and Geography, Faculty of Biology, University of Warsaw, Warsaw, Poland. 4 Present address: USDA, NCAUR, BFPM, 1815 North University, Office 3133, Peoria, Illinois 61604 U. S. A. 5 Author for correspondence ([email protected])

Communicating Editor: Allan J. Bornstein

Abstract— The genus Lilaeopsis (Apiaceae subfamily Apioideae) comprises 15 and exhibits both American amphitropic and amphi- antarctic patterns of disjunction. The group is difficult taxonomically because of its simplified habit, phenotypic plasticity of vegetative charac- ters, and extensive variation in fruit characters. Sequence data from the nrDNA ITS and cpDNA rps16 intron and rps16-trnK intergenic spacer regions were obtained for 60 accessions, representing 13 species of Lilaeopsis and five closely related outgroup genera from the North American Endemics clade of tribe Oenantheae. These molecular data were subjected to maximum parsimony, Bayesian inference, and dispersal-vicariance analyses in an effort to reconstruct evolutionary relationships and infer biogeographic scenarios. The results suggest that: (1) L. macloviana , L. masonii, and L. occidentalis, distributed in western South America and western North America, collectively represent a single, polymorphic species of amphitropic distribution; (2) The Australasian species L. brisbanica, L. novae-zelandiae, L. polyantha, and L. ruthiana comprise a well- supported clade. However, L. novae-zelandiae is not monophyletic, but may be rendered so by the inclusion of all Australasian taxa into one polymorphic species; (3) L. mauritiana from Mauritius is closely related to L. brasiliensis from South America and may even be subsumed under the latter pending further investigation; and (4) Lilaeopsis probably originated in South America following a dispersal of its ancestor from North America. A minimum of seven dispersal events is necessary to explain its present-day distribution, including one dispersal from South America to Australia or New Zealand, two dispersals between Australia and New Zealand, and three dispersals from South America to North America. Keywords— Amphitropic disjunction , Apioideae, cpDNA , ITS , phylogeny.

The genus Lilaeopsis Greene (Apiaceae subfamily Apioideae) nized by Hill (1928) were consolidated within the single poly- consists of small, creeping, rhizomatous, perennial herbs morphic taxon L. novae-zelandiae ; and three additional species occupying damp, marshy, or truly aquatic habitats, with fif- from Australia were interpreted as intergrading forms of teen species and four infraspecific taxa currently recognized L. polyantha . In general, vegetative characters were deemed of ( Table 1 ). The group is difficult taxonomically because the little value for distinguishing among species, and fruit char- have a simplified and generally similar vegetative acters of the three polymorphic species were observed to be morphology. In all species the are linear, hollow, and extremely variable, sometimes showing overlapping varia- transversely septate, being derived from the rachis-axis of a tion between species. To confound matters, specimens are compound , with the septae corresponding to the posi- often impossible to identify to species without mature fruits, tions of pinnae insertion in pinnatifid leaves ( Kaplan 1970 ; so in some instances geography must be used to circumscribe Charlton 1992). These rachis leaves are different from those them. As stated by Affolter (1985) , “Morphological simplifi- typical of other umbellifers (such as carrot, dill, parsnip, and cation has reduced the number of characters available to the celery) where they are generally pinnately compound and taxonomist; phenotypic plasticity has diminished the utility dissected. of the few characters which remain.” Further consolidation The taxonomic history of Lilaeopsis is detailed by Affolter of species was also considered by Affolter (1985) , but “would (1985) and underscores the problems in classifying plants carry the process a step too far and would obscure real dif- having a reduced vegetative morphology. Affinities to both ferences that exist among the South American, New Zealand, Araliaceae (i.e. Hydrocotyle L.) and Apiaceae have been pro- and Australian populations.” posed, with the phylogenetic position and monophyly of Molecular systematic studies indicate that Lilaeopsis is a Lilaeopsis only confirmed recently (Downie et al. 2000, 2001 , member of Apiaceae tribe Oenantheae ( Downie et al. 2000 ). 2008 ; Petersen et al. 2002 ). Through extensive field inves- While no unique morphological synapomorphy supports the tigations, cultivation of living material of most species in a monophyly of Oenantheae (Petersen et al. 2002), the plants do common-garden coupled with the growth of many collec- share certain traits, such as glabrous stems and leaves, clus- tions under different environmental treatments, and sta- ters of fibrous or tuberous-thickened roots, globose to broadly tistical analyses, Affolter (1985) showed that the shape and ovate spongy-thickened fruits, and a preference for wet habi- size of the rachis leaves of Lilaeopsis are readily modified in tats ( Downie et al. 2008 ). The “spongy cells” within the fruits response to various degrees of submergence and light inten- are storage tracheids, nearly isodiametric in form and bigger sity. Fruit characters, so important in umbellifer classifica- in diameter than typical tracheids. They enhance buoyancy tion, continued to receive emphasis, but recognition of their and facilitate dispersal in aquatic environments when the extensive variation in some taxa resulted in merging many fruits dry out and these cells become filled with air (Briquet previously recognized species. As examples, six previously 1897 ; Hill 1927 ; Affolter 1985 ). Like Lilaeopsis, some plants of described species and several infraspecific taxa were consol- the tribe possess rachis or rachis-like leaves, such as Oxypolis idated within the single polymorphic species L. macloviana Raf., Ptilimnium Raf., Cynosciadium DC., and Limnosciadium from South America; the three New Zealand species recog- Mathias & Constance. The genus Lilaeopsis is unequivocally

789 790 SYSTEMATIC BOTANY [Volume 36

Table 1. Lilaeopsis taxa and their geographic distributions (after of Lilaeopsis have also been collected from the Kerguelen Affolter 1985 , Bean 1997 , and Petersen and Affolter 1999 ). Asterisks denote Archipelago ( Affolter 1985 ). taxa not included in the molecular analysis. Both northern and southern hemispheric origins for

Taxon Distribution Lilaeopsis have been suggested. Dawson (1971), for example, indicated that Lilaeopsis probably originated in the northern L. attenuata (Hook. & Arn.) Fern. Argentina subsp. attenuata hemisphere, whereas Hill (1929) postulated an Antarctic ori- L. attenuata subsp. ulei (Pérez-Mor.) Brazil gin, with two major subsequent migration routes northward Affolter* (one route leading to New Zealand and Australia, and the L. brasiliensis (Glaz.) Affolter Argentina, Brazil, Paraguay other leading into and through the Andes of South America L. brisbanica A. R. Bean Australia into North America). Raynal (1977) described Lilaeopsis as a L. carolinensis J. M. Coult. & Rose U. S. A. (Atlantic and Gulf coasts) “genuinely Gondwanian genus.” The southern hemisphere South America (Argentina, Brazil, Paraguay, Uruguay) has traditionally been considered to exhibit a vicariant his- Europe (Portugal, Spain) tory, with the fragmentation of Gondwana leading to the L. chinensis (L.) Kuntze Canada and U. S. A. (Atlantic and division of its ancestral biota (Sanmartín and Ronquist 2004). Gulf coasts) However, recent biogeographic studies based on molecular L. fistulosa A. W. Hill* southeastern Australia dating estimates have revealed that many plant transoce- L. macloviana (Gand.) A. W. Hill western South America (Colombia to Tierra del Fuego, Falkland anic disjunctions are relatively recent, thus better explained Islands) by long-distance dispersal rather than by vicariance (Les L. masonii Mathias & Constance U. S. A. (California) et al. 2003 ; Sanmartín and Ronquist 2004 ; de Queiroz 2005 ). L. mauritiana G. Petersen & Affolter Mauritius, Madagascar In explaining the disjunction of L. carolinensis in North and L. novae-zelandiae (Gand.) A. W. Hill New Zealand South America, Affolter (1985) speculated that its origin was L. occidentalis J. M. Coult. & Rose Canada and U. S. A. (Pacific coast) probably in South America, with subsequent introduction via L. polyantha (Gand.) H. Eichler southeastern Australia L. ruthiana Affolter New Zealand long-distance dispersal by migrating birds to North America. L. schaffneriana (Schltdl.) U. S. A. (), Such intercontinental dispersal by birds is regarded as a via- J. M. Coult. & Rose subsp. (Sonora) ble explanation for widely disjunct aquatic plant distributions recurva (A. W. Hill) Affolter ( Raven 1963 ; Les et al. 2003 ). However, in explaining amphi- L. schaffneriana subsp. schaffneriana Mexico, northwestern South tropic distribution patterns in other genera of Apiaceae and America L. tenuis A. W. Hill* Brazil taxa from other families having transoceanic disjunctions in the southern hemisphere, the predominant direction of dis- persal appears to have been from the northern to southern monophyletic. All species are true aquatics or occupy wet hemisphere, even for those genera that are much diversified habitats. They are all tiny, perennial, rhizomatous herbs, with in the southern hemisphere (reviewed in Spalik et al. 2010). linear to spatulate, simple, septate rachis leaves. The inflo- A phylogenetic hypothesis for Lilaeopsis is necessary to eluci- rescences are few-flowered, the are simple, and the date its historical biogeography, specifically its place of origin schizocarp lacks a free carpophore. Such attributes are rare in and direction(s) of long-distance dispersal. the subfamily Apioideae and, when considered collectively, The major objectives of this study are to provide an esti- unambiguously circumscribe the genus. The results of molec- mate of species-level relationships within the genus Lilaeopsis , ular systematic studies also demonstrate strong support for especially among its New World members, and to better its monophyly (Petersen et al. 2002; Downie et al. 2008); inter- understand the historical biogeographic processes that have specific relationships within Lilaeopsis , however, have yet to shaped its biodiversity. Since most morphological characters be elucidated. of Lilaeopsis are either reduced relative to other umbellifers or Lilaeopsis combines both American amphitropic (or bipolar) readily modified by the environment, they offer relatively lit- and amphiantarctic (or amphi-South Pacific) patterns of dis- tle data for phylogeny estimation. Therefore, molecular data junction ( Affolter 1985 ; Table 1 ). Lilaeopsis occurs in the tem- are essential. To reach these objectives, we analyze the nuclear perate zones of both North and South America, with two taxa ribosomal ITS and cpDNA rps16 intron and rps16 - trnK inter- extending into the tropics at high elevations: L. macloviana genic spacer regions, as previous studies have demonstrated exists along the Andes from Colombia south to Tierra del the utility of these loci in resolving interspecific relationships Fuego; and L. schaffneriana subsp. schaffneriana is distributed in Apiaceae tribe Oenantheae (Petersen et al. 2002; Downie et in Mexico and the Andes of northwestern South America. al. 2008 ). Lilaeopsis was introduced in the Iberian Peninsula, where it is now naturalized (Affolter 1985); these plants have been referred to as either L. carolinensis (Affolter 1985) or L. atten- Materials and Methods uata ( Tutin 2001 ; Almeida and Freitas 2001 ). In the southern Leaf material for DNA extraction was obtained primarily from voucher hemisphere, Lilaeopsis is widely disjunct, occurring predomi- specimens prepared by Affolter during the course of his monographic nantly in cool temperate regions of South America, Australia, study of Lilaeopsis ( Affolter 1985 ). Additional material was obtained and New Zealand, with isolated populations occurring in the from other herbarium specimens (AK, ARIZ, ASU, GA, ILL, ILLS, ISU, LSU, MICH, MO, TEX, and UC) and through fieldwork in southeastern southwestern and southern Indian Ocean. On the island of Arizona. Four species of Lilaeopsis are sold in the commercial aquarium Mauritius, L. mauritiana is known from only a single local- trade ( L. brasiliensis, L. macloviana, L. mauritiana, and L. novae-zelandiae ) ity and may be a local endemic (Petersen and Affolter 1999). and are commonly marketed as water-, water chives, microsword, Lilaeopsis has been reported from Madagascar based on a and grasswort; therefore, material of these species (five accessions) was obtained from two aquarium supply companies (Tropica Aquarium Plants, single, sterile specimen (Raynal 1977; Affolter 1985) that has Denmark, and Florida Aquatic Nurseries, Florida). For four other acces- been identified provisionally as “L. aff. mauritiana G. Petersen sions, DNA and leaf material were supplied to us directly (G. Petersen, & Affolter?” (Sales et al. 2004). Sterile, unidentified specimens University of Copenhagen, Denmark). In total, 13 species of Lilaeopsis 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 791 were represented by 54 accessions. Three taxa ( L. attenuata subsp. ulei , L. fistulosa, and L. tenuis) were not included in the molecular study for lack of adequate material: Lilaeopsis attenuata subsp. ulei is restricted to Serra do Itatiaia, Brazil; L. fistulosa is known only from a few locations in Eastern New South Wales, Australia; and L. tenuis is known only from two collections from southeastern Brazil ( Affolter 1985 ). Source and voucher information for all accessions of Lilaeopsis are presented in Appendix 1. As outgroups, we included representatives of five closely related genera (six accessions) of Apiaceae tribe Oenantheae (Atrema DC., Cynosciadium, Neogoezia Hemsl., Ptilimnium, and Trepocarpus Nutt. ex DC.; Appendix 1). Previous phylogenetic studies have indicated that Lilaeopsis is a member of the North American Endemics clade of tribe Oenantheae, along with these five genera plus Daucosma Engelm. & A. Gray ex A. Gray, Limnosciadium , and Oxypolis ( Hardway et al. 2004 ; Downie et al. 2008 ). This clade represents a group of taxa that is primarily distributed in North America, but also includes Lilaeopsis with a broader distribution. All phy- logenetic trees were rooted with Atrema, Neogoezia , and Trepocarpus , as the aforementioned phylogenetic studies have revealed that these three gen- era comprise a clade that is more distantly related to Lilaeopsis than either Cynosciadium or Ptilimnium . Total genomic DNA was extracted using a DNeasy plant mini kit (QIAGEN, Valencia, California), using 12–22 mg of dried leaf mate- rial following experimental strategies described elsewhere (Downie and Katz-Downie 1996, 1999 ; Lee and Downie 2006; Spalik and Downie Fig. 1. Map of the 1.8 kb locus of Lilaeopsis cpDNA showing the posi- 2006 ). Amplification of the entire ITS region (ITS1, 5.8S rDNA, ITS2) was tions of genes rps16 and trnK(UUU) 5 ′exon. The gene rps16 is interrupted attained using primers 18S-for (Feist and Downie 2008) and C26A (Wen by an intron; an intergenic spacer separates gene rps16 from gene 5′trnK . and Zimmer 1996). For some accessions the two spacer regions were each The two cpDNA regions sequenced in the phylogenetic analyses are amplified separately using the following primer pairs: 18S-ITS1-F and indicated by brackets. The arrows represent the directions and approxi- 5.8S-ITS1-R for ITS1, and ITS-3N and C26A for ITS2 (Spalik and Downie mate positions of the primers used in PCR amplifications and/or DNA 2006 ). The ITS sequences were obtained for 57 accessions of Lilaeopsis sequencing. Fourteen primers, labeled L1 through L14, were devel- and outgroups. Amplifications of the rps16 intron and rps16 -trnK inter- oped specifically for this study; their sequences, written 5′ to 3′, are pre- genic spacer regions were obtained for most accessions using primer pair sented below. The eight remaining primers have been used in previous 5′exon(rps16) and 3′exon(rps16) for the intron and primer pairs rps16C studies of Apiaceae tribe Oenantheae and their details are presented in and trnK or rps16–2 and trnK for the intergenic spacer region ( Downie Downie et al. (2008). L1 : TATC(G/A)C(G/A)CGGGAATC(G/T)A(C/G) et al. 2008 ; Fig. 1 ). For those accessions where the spacer region could CGTT(C/T)A; L2 : CTTTCTCTTCGGGATCGAACATCA; L3 : AAAGCAA not be amplified with these primers, three internal primers were used CGTGCGACTTGAAGGAC; L4 : TAAACGCTCGATTCCCG(C/T)G(C/T) (3′exon-1, trnK-1R, and trnK-1; Downie et al. 2008 ). In addition, 14 prim- GATA; L5 : TCAAAGTGTATCGCACGGGAATCG; L6 : TCATTTGTACCC ers (L1-L14) were designed specifically for this study in order to amplify ATAACTCAAGTTGG; L7 : TCCAACTTGAGTTATGGGTACAAATG; L8 : or sequence both noncoding regions (primer sequences are provided in AGCGGGAACGTTTAAATAACTTTGA; L9 : AAGGAAGAGATCTTCGG Fig. 1 ). For some accessions, primer pairs L3 and L2 or L3 and 3′exon(rps16) AACGTGG; L10 : AACCGACCAAATGAAGGAACTC; L11 : TGGGAGTT were used to obtain intron data; however, because primer L3 overlaps the CCTTCAATTGGTCG; L12 : TTTGTTCGATACACTGTTGTCA; L13 : TGTA 5′exon/intron boundary, the first 30 bp of sequence at the 5′ end of the GTGCCAATCCAACACAAGCC; L14 : AGAAATGTCAAATTTATAGACC intron could not be obtained. The cpDNA rps16 intron and rps16 - trnK ACCTCTTAG. sequence data were obtained for 40 accessions of Lilaeopsis and outgroups, with 37 accessions common to both ITS and cpDNA studies. In general, DNA extractions from herbarium specimens, particularly those more than missing data. Unambiguous alignment gaps were scored as presence/ 20 yrs old, resulted in poor or no PCR products. Many of these DNAs absence characters using the simple indel coding method of Simmons were amplified repeatedly, under different stringency conditions and and Ochoterena (2000). Gaps of equal length in more than one sequence using varying quantities of DNA, but failed to work. Of 135 DNA extrac- were coded as the same presence or absence character state if they could tions attempted, only about half of these yielded PCR products of suf- not be interpreted as different duplication or insertion events. Indels of ficient concentration to be sequenced and included in the study. Technical similar location, but with different lengths, were coded as different binary difficulties in working with Lilaeopsis DNAs have also been reported by characters. Bootstrap values were calculated from 100 replicate analyses Fiedler et al. (in press). All ITS and cpDNA sequences obtained have been using TBR branch swapping and simple stepwise addition of taxa. Bremer deposited in GenBank (Appendix 1). (1994) support values were calculated using TreeRot version 3 ( Sorenson Nucleotide sequences of the ITS and cpDNA regions were each aligned and Franzosa 2007). Prior to combining the ITS and cpDNA data, the initially using the default pairwise and multiple alignment parameters in incongruence length difference (ILD) test of Farris et al. (1995) was carried Clustal X (gap opening cost = 15.00; gap extension cost = 6.66; DNA transi- out using the partition-homogeneity test of PAUP* to examine the extent tion weight = 0.50; Jeanmougin et al. 1998), then rechecked and adjusted of conflict between datasets. This test was executed with 1,000 replicate manually as necessary. Gaps were positioned to minimize nucleotide mis- analyses, using the heuristic search option, simple stepwise addition of matches. Ambiguously aligned regions were excluded from the analysis. taxa, and TBR branch swapping. These aligned ITS and cpDNA data matrices are available in TreeBASE The BI analysis was implemented using MrBayes version 3.1.2 (study number S11209), as well as in Bone (2007). Characteristics of the (Ronquist and Huelsenbeck 2003). Prior to analysis, MrModeltest version aligned ITS and cpDNA sequences, separately and combined, were 2.2 (Nylander 2004) was used to select an evolutionary model of nucle- obtained with uncorrected pairwise nucleotide distances calculated in otide substitution that best fits these data, as selected by the AIC estima- PAUP* version 4.0b10 ( Swofford 2002 ). The percentage of data matrix tor. The settings appropriate for this best-fit model (SYM + I for ITS and cells scored as missing data in the ITS and cpDNA matrices was 0.8% and GTR + G for cpDNA) were put into a MrBayes block in PAUP* (nst = 1.5%, respectively. These regions represented a portion of the 5.8S gene 6, rates = propinv and nst = 6, rates = gamma, respectively). In the BI (for those accessions where the two spacer regions had to be amplified analysis of combined data (with scored gaps), both models of nucleotide separately) and several small areas within the cpDNA intron and inter- substitution were used for their respective molecular partitions, while genic spacer that could not be sequenced with the primers at hand. a standard model for unordered characters was used for gap data. The The three data matrices (ITS, cpDNA, and combined ITS/cpDNA) priors on state frequencies, and rates and variation across sites were esti- were each analyzed using maximum parsimony (MP) and Bayesian mated automatically from the data assuming no prior knowledge about inference (BI) methods. In the MP analysis implemented using PAUP*, their values. Two independent analyses were each run simultaneously characters were treated as unordered and all character transformations for one million generations, with tree sampling occurring every 100 gen- were equally weighted. Heuristic MP searches were replicated 1,000 times erations. Starting trees were chosen at random. One thousand trees were with random stepwise addition of taxa, tree-bisection-reconnection (TBR) discarded (as “burn-in”) before stationarity was reached, prior to deter- branch swapping, and saving multiple trees. Gap states were treated as mining the posterior probability (PP) values from the remaining trees. 792 SYSTEMATIC BOTANY [Volume 36

The biogeographic history of Lilaeopsis was reconstructed using sim- those between L. brasiliensis and L. mauritiana accessions also plified, fully resolved trees of species relationships obtained from analy- ranged from 0.33% to 1.0%. The highest levels of infraspecific ses of combined ITS/cpDNA data. The differences between these trees were in the ancestral distributions assigned to species L. carolinensis and sequence divergence (2.7%) occurred between accessions of L. schaffneriana. Each terminal represented a major lineage supported L. novae-zelandiae . Ranges for other infraspecific divergence by these phylogenetic results; however, three species (L. mauritiana, values were 0–1.2% for the four accessions of L. macloviana , L. macloviana, and L. masonii) were not included as terminals in the biogeo- 0–0.34% for the seven accessions of L. occidentalis , 0–1.2% graphic analysis: L. mauritiana arises from within a paraphyletic L. brasil- for the five accessions of L. brasiliensis, 0–0.33% for the three iensis ; and L. macloviana and L. masonii are subsumed within L. occidentalis . Outgroups were the same as included in the phylogenetic study. Four unit accessions of L. mauritiana, and 0–0.17% for the ten accessions areas were defined: (A) North America (Canada, U. S. A., and Mexico); of L. schaffneriana . The two accessions of L. schaffneriana subsp. (B) South America; (C) Australia; and (D) New Zealand. The occurrence of schaffneriana from Mexico had identical ITS sequences to two Lilaeopsis in Europe (Spain and Portugal) is recent (Affolter 1985; Almeida accessions of L. schaffneriana subsp. recurva from Arizona. and Freitas 2001 ), thus this region was not included as an ancestral area. Similarly, only few reports of dubiously identified material of Lilaeopsis The MP analysis of the ITS data matrix recovered 15 min- from the Dominican Republic, Kerguelen Islands, and Madagascar pre- imal length 452-step trees (CI = 0.757 and 0.690, with and cluded their inclusion in the biogeographic study. The dispersal-vicari- without uninformative characters, respectively; RI = 0.918). ance analysis was carried out using DIVA version 1.1, which reconstructs The strict consensus of these trees is presented in Fig. 2 , with the distribution of ancestral areas that minimize dispersal events under a accompanying bootstrap and Bremer support values. The MP parsimony criterion (Ronquist 1996, 1997 ). Two optimizations were per- formed: first, with an unconstrained number of unit areas for each ances- analysis of the ITS matrix plus 28 binary-scored indel charac- tral node and second, with this number restricted to two areas ( Ronquist ters recovered 15 trees, each of 488 steps (CI = 0.758 and 0.698, 1996 , 1997 ; Downie et al. 2008 ). The rationale for the second optimization with and without uninformative characters, respectively; RI = is that in an unconstrained analysis, the ancestral distribution at or near 0.920). The strict consensus of these trees (not shown) differed the base of the tree may be inferred to be widespread and include most or all individual unit areas inhabited by the terminals because of uncertainty from that presented in Fig. 2 by showing a monophyletic (Ronquist 1996). Because we were interested in inferring the ancestral dis- L. brasiliensis arising from within a paraphyletic L. mauritiana , tributions if the group had a more restricted (and likely realistic) distribu- L. macloviana accessions 2925 and 2518 comprising a clade, and tion, we repeated the analysis by limiting the number of ancestral areas P. nuttallii 2405 and C. digitatum 1804 also comprising a clade. assigned to each node to two. Bootstrap support values for identical clades were similar in both analyses. To reveal the distribution of indels through- Results out the phylogeny and to show relative branch lengths, the pattern of indel distribution was mapped onto a single, arbi- ITS Analysis—Sequence characteristics of the ITS region trarily selected tree derived from MP analysis of ITS sequences are provided in Table 2 . The length of the entire ITS region and scored gaps (Fig. 3). Patterns of indel distribution sup- across 57 accessions ranged from 594–608 bp, and align- port the monophyly of Lilaeopsis and several species or spe- ment of these sequences resulted in a matrix of 634 positions, cies groups. A single unique indel supports the monophyly including 41 unambiguous gaps. Twenty-eight of these gaps of the Australasian species L. brisbanica, L. novae-zelandiae, were parsimony informative, with all but one of them one to L. polyantha, and L. ruthiana, and four additional synapomor- three bp in size. The largest gap, of eight bp, characterized phic indels support the subsequent major dichotomy in this all accessions of L. attenuata, L. carolinensis, L. chinensis , and species group. Two synapomorphic indels support the branch L. schaffneriana. The number of parsimony informative posi- leading to L. carolinensis, L. attenuata, L. schaffneriana , and tions was 179, with the ITS-2 region containing more infor- L. chinensis . is the only species supported mative positions (95) than either ITS-1 (76) or 5.8S (eight). by a uniquely occurring indel. Four indels were homoplas- Uncorrected pairwise ITS sequence divergence values ranged tic, each occurring 2–4 times on the tree. With the exceptions from identity (for several conspecific taxa) to 23.7% (between of a slightly greater resolution among the ten included acces- L. brasiliensis and P. nuttallii ). Within Lilaeopsis , ITS sequence sions of L. schaffneriana and the union of outgroups P. nuttallii divergence values ranged from identity to 9.6% (the latter and C. digitatum into a weakly-supported clade, the majority- between L. occidentalis and L. chinensis). Pairwise sequence rule consensus tree derived from the BI analysis was topo- comparisons between L. occidentalis and L. masonii accessions logically identical to that tree inferred through MP without ranged from identity to 0.17%, those between L. occidentalis scored gaps. The PP values are presented in Fig. 2 for those and L. macloviana accessions ranged from 0.33% to 1.0%, and nodes occurring in both MP and BI trees. As a result of phylogenetic analyses of ITS data, seven major Table 2. Sequence characteristics of the ITS and cpDNA (rps16 intron clades are circumscribed within Lilaeopsis that correspond to and rps 16 -trnK intergenic spacer) regions, separately and combined. species or species groups. These include: (1) L. carolinensis , (2) L. attenuata , (3) L. schaffneriana , (4) L. chinensis , (5) L. brasilien- Sequence characteristic ITS cpDNA Combined sis and L. mauritiana species group, (6) L. brisbanica, L. novae- No. of terminals 57 40 37 zelandiae , L. polyantha , and L. ruthiana species group, and (7) Length variation (range in bp) 594–608 1,506–1,566 2,113–2,172 L. macloviana, L. masonii, and L. occidentalis species group. The No. of aligned positions 634 1,630 2,264 first four of these major clades comprised a well-supported No. of excluded positions 0 40 40 monophyletic group in all analyses (93% BS, 1.00 PP), with No. of constant positions 372 1,436 1,851 No. of autapomorphic positions 83 66 129 this group characterized by eight bp and one bp alignment No. of parsimony informative 179 88 244 deletions. The clade of L. macloviana, L. masonii, and L. occiden- positions talis was a sister group to the clade comprising all remaining No. of unambiguous gaps 41 47 56 Lilaeopsis. With the exception of the L. carolinensis clade, which No. of parsimony informative gaps 28 28 45 was supported weakly in both MP and BI analyses (64% BS, Max. pairwise sequence divergence 0.69 PP), support for the remaining clades comprising three or (%) (Lilaeopsis only/all accessions) 9.6/23.7 1.7/4.6 3.8/9.4 more accessions was moderate to high (BS values ≥ 75% and 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 793

Fig . 2. Strict consensus of 15 minimal length 452-step trees derived from maximum parsimony (MP) analysis of 57 nrDNA ITS sequences obtained from 13 species of Lilaeopsis and 5 outgroups (CI = 0.757 and 0.690, with and without uninformative characters, respectively; RI = 0.918). The majority-rule consensus tree derived from Bayesian inference analysis of these data was topologically identical to this MP strict consensus tree. Numbers above nodes represent MP bootstrap and BI posterior probability values, respectively. Numbers below nodes represent Bremer support values.

PP values ≥ 0.84). However, not all species were monophyl- Furthermore, two accessions of L. macloviana from South etic in all analyses. was monophyletic America (2518 and 2925) were basal within this clade relative when gaps were included as additional characters in the MP to North American L. occidentalis and L. masonii and the two analysis ( Fig. 3 ), but not when gaps were excluded or in the BI other accessions of L. macloviana from South America. analysis. In all trees, L. mauritiana was paraphyletic relative to cpDNA Analysis— Sequence characteristics of the cpDNA L. brasiliensis . Lilaeopsis novae-zelandiae comprised two major noncoding regions are provided in Table 2 . The concatenated lineages, with the single included accession of L. ruthiana a rps16 intron and rps16 - trnK intergenic spacer regions for sister group to one of these lineages, and the clade of L. bris- 40 included accessions varied in length from 1,506 bp ( L. brasil- banica and L. polyantha a sister group to the other. Within the iensis ) to 1,566 bp ( L. occidentalis ). Alignment of these sequences L. macloviana, L. masonii and L. occidentalis clade, neither L. occi- resulted in a matrix of 1,630 positions, with 47 unambiguous dentalis nor L. macloviana comprised a monophyletic group. gaps. Twenty-eight of these gaps were parsimony informative, 794 SYSTEMATIC BOTANY [Volume 36

Fig . 3. One of 15 minimal length 488-step trees derived from maximum parsimony analysis of 57 nrDNA ITS sequences and 28 binary-scored infor- mative gaps obtained from 13 species of Lilaeopsis and five outgroups (CI = 0.758 and 0.698, with and without uninformative characters, respectively; RI = 0.920). The pattern of indel distribution is indicated (solid bar, synapomorphic indel; open bar, homoplastic indel). Scale bar indicates numbers of inferred changes along each branch.

with 16 of them a single bp in size. The remaining informa- several conspecific taxa to 4.6% (between L. carolinensis and tive gaps ranged between two and 26 bp, with the largest P. nuttallii ); among Lilaeopsis cpDNAs, maximum pairwise gap occurring only in Trepocarpus and Atrema . The next larg- interspecific sequence divergence values approached 1.7% est informative gaps, of 18, 14, and 11 bp, supported other (between L. occidentalis and L. carolinensis ). These values were outgroup lineages. Upon the exclusion of 40 ambiguously much lower than those of the ITS region; indeed, for the same aligned positions, the number of variable positions was 154, pairwise comparisons, sequence divergence values for the with the rps16 - trnK intergenic spacer region having a greater ITS region were four to five times higher than those values number of variable positions (85) than that of the intron (69). inferred for cpDNA. Similarly, relative to the length of their Uncorrected pairwise sequence divergence values across all respective unambiguous alignments, the proportion of par- ingroup and outgroup accessions ranged from identity for simony informative nucleotide positions was approximately 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 795 five times greater for ITS (28%) than for cpDNA (6%). The consistent with that presented in Fig. 4. Differences between number of parsimony informative gaps, however, was iden- them included the clade of L. brasiliensis 2989 and L. brasilien- tical for both ITS and cpDNA data matrices (28). Pairwise sis 3179, the placement of L. brasiliensis 2719 in a clade with cpDNA sequence divergence estimates for L. schaffneriana L. brasiliensis 2660 and all four accessions of L. carolinensis , ranged from identity to 0.14%, with the single accession of and the placement of L. macloviana 2925 as one branch of a tri- L. schaffneriana subsp. schaffneriana from Mexico having chotomy along with the clade of all three accessions of L. occi- an identical cpDNA sequence to those of four accessions of dentalis and the clade of all remaining accessions of Lilaeopsis . L. schaffneriana subsp. recurva from Arizona. Pairwise compari- Patterns of indel distribution (not shown) are similar to those sons between L. brasiliensis and L. mauritiana accessions ranged exhibited by ITS data. Nine indels were homoplastic, each from identity to 0.54%, and those between L. occidentalis and occurring two to four times on a single, arbitrarily selected L. macloviana accessions ranged between 0.68% and 0.86%. cpDNA derived tree. The majority of indels supported the The MP analysis of the cpDNA data matrix recovered 26 monophyly of Lilaeopsis and the outgroup lineages. Lilaeopsis minimal length 185-step trees (CI = 0.887 and 0.824, with and chinensis and L. occidentalis were the only species represented without uninformative characters, respectively; RI = 0.934). by two or more accessions that were supported by uniquely The strict consensus of these trees is presented in Fig. 4 , occurring indels. Lilaeopsis carolinensis was supported by a with accompanying bootstrap and Bremer support values. single homoplastic indel. The BI majority-rule consensus tree The MP analysis of the cpDNA data matrix plus 28 binary- was almost fully congruent to that inferred using MP (omit- scored indel characters recovered 36 trees, each of 224 steps ting scored gaps), with the only exception of L. macloviana (CI = 0.857; CI excluding uninformative characters = 0.798; 2925 being a sister group to a clade comprising all other acces- RI = 0.928). The strict consensus of these trees (not shown) was sions of Lilaeopsis (including L. occidentalis ).

Fig . 4. Strict consensus of 26 minimal length 185-step trees derived from maximum parsimony (MP) analysis of 40 cpDNA rps16 intron and rps16 - trnK intergenic spacer sequences representing 12 species of Lilaeopsis and five outgroups (CI = 0.887 and 0.824, with and without uninformative charac- ters, respectively; RI = 0.934). The majority-rule consensus tree derived from BI analysis of these data was nearly topologically congruent to this MP strict consensus tree. Numbers above nodes represent MP bootstrap and BI posterior probability values, respectively. Numbers below nodes represent Bremer support values. 796 SYSTEMATIC BOTANY [Volume 36

The trees obtained through MP and BI analyses of cpDNA sions were available, pairwise infraspecific sequence diver- sequences were less resolved and their branches generally gence values were as follows: L. brasiliensis (0.05–0.55%); more poorly supported than those inferred using ITS data. L. carolinensis (0–0.10%); L. chinensis (0.10%); L. mauriti- The previously delimited L. carolinensis, L. attenuata , and ana (0.05–0.19%); L. novae-zelandiae (0–1.3%); L. occidentalis L. chinensis clades along with the L. brasiliensis and L. mau- (0–0.15%); and L. schaffneriana (0–0.10%). DNA data from the ritiana species group comprised a large polytomy that was single included accession of L. schaffneriana subsp. schaffneri- supported weakly (61% BS, 0.90 PP, and a Bremer sup- ana were identical to that from one accession of L. schaffneriana port value of 1) in the cpDNA-derived trees. This large subsp. recurva . clade was a well-supported sister group to L. schaffneriana . The MP analysis of the combined ITS and cpDNA matrix The previously delimited L. brisbanica, L. novae-zelandiae , (excluding scored gaps) resulted in 12 minimal-length trees of L. polyantha, and L. ruthiana species group comprised two lin- 541 steps each (CI = 0.810 and 0.747, with and without unin- eages basal to the clade of all aforementioned taxa, with the formative characters, respectively; RI = 0.900), and the strict single accession of L. macloviana and a clade of three acces- consensus of these trees, with accompanying branch sup- sions of L. occidentalis comprising successively sister branch- port values, is presented in Fig. 5. Repeating the MP analysis ing lineages with ITS. The separation of L. occidentalis from with the 45 informative gaps scored as additional characters L. macloviana , however, is poorly supported. Major differ- resulted in trees whose strict consensus was almost topologi- ences between the ITS and cpDNA trees include the relative cally identical to that inferred without scored gaps (CI = 0.794 position of L. macloviana , the placement of L. brisbanica and and 0.736, with and without uninformative characters; RI = L. polyantha relative to the two lineages of L. novae-zelandiae , 0.898). The only substantive difference between these strict and collapse of the L. brasiliensis and L. mauritiana species consensus trees was a monophyletic L. mauritiana arising from group. within a paraphyletic L. brasiliensis when gaps were included Combined ITS and cpDNA Analysis— Results of a parti- as additional characters (Fig. 5, inset); however, support for tion-homogeneity test on a set of 37 accessions common to the monophyly of L. mauritiana was weak. The majority-rule both ITS and cpDNA analyses revealed that these data matri- consensus tree derived from the BI analysis (excluding scored ces yield significantly incongruent phylogenetic estimates gaps) was almost identical to the MP strict consensus tree (p = 0.01). We acknowledge, however, that serious doubts inferred without scored gap characters (PP values are pre- have been raised regarding the value of this test as a criterion sented in Fig. 5 ), with the only difference being the union of for deciding whether data should be combined into a single L. brisbanica and L. polyantha as a weakly supported clade in phylogenetic analysis (Yoder et al. 2001; Barker and Lutzoni the former. The BI majority-rule consensus tree obtained from 2002) and, in the absence of information suggesting that past all data (including scored gaps) is presented in Fig. 6A , while hybridization events or other factors may have caused this Fig. 6B presents a single tree derived from these same data. discordance, combine all molecular data into a single data The only major difference between the results of the MP and matrix for simultaneous consideration. Such an analysis BI analyses of all data is the relative position of L. chinensis . also provided the greatest number of relationships receiving Phylogenetic analyses of the combined data set resulted in strong support. Ptilimnium nuttallii was not included in the better resolved trees than either of the partitioned analyses. combined analysis because different accessions were used in The combined data set also provided the greatest number of the partitioned analyses, although the species is unequivo- relationships receiving strong support. On the basis of these cally monophyletic ( Feist and Downie 2008 ). Lilaeopsis masonii combined data, seven major clades are confirmed within was also not included in the combined analysis because Lilaeopsis : (1) L. carolinensis , (2) L. attenuata , (3) L. schaffneri- cpDNA data were unavailable. Alignment of ITS and cpDNA ana, (4) L. chinensis, (5) L. brasiliensis and L. mauritiana species sequences for 33 accessions of Lilaeopsis and four outgroup group, (6) L. brisbanica, L. novae-zelandiae , L. polyantha , and taxa resulted in a matrix of 2,264 positions, with 40 positions L. ruthiana species group, and (7) L. occidentalis and L. maclo- excluded from subsequent analyses because of alignment viana species group (L. masonii was not included in the com- ambiguity ( Table 2 ). Of the remaining positions, 373 were bined analysis). Support for each of the clades comprising variable and 244 of these were parsimony informative. Forty- two or more taxa was high (BS ≥ 95%, PP ≥ 0.95, Bremer sup- five of 56 unambiguous alignment gaps were also parsimony port 5–9 steps). Lilaeopsis macloviana and L. occidentalis , distrib- informative. Uncorrected pairwise sequence divergence val- uted in western South America and western North America, ues across all accessions ranged from identity to 9.4%, the respectively, collectively comprise a well-supported clade sis- latter between L. brasiliensis and the outgroup Trepocarpus ter group to all other members of the genus. Lilaeopsis maclo- aethusae. Within Lilaeopsis , interspecific sequence variation viana and L. occidentalis also represent sister taxa, although only ranged from 0.09% (between L. ruthiana and L. novae-zelandiae ) one accession of L. macloviana was included in the combined to 3.8% (between L. occidentalis and L. chinensis). The next low- analysis. Combined sequence divergence estimates between est values in interspecific comparisons were between L. brasil- these two species, however, were not high (0.74–0.86%), and iensis and L. mauritiana (0.15–0.48%), between L. brisbanica and within the range of those values obtained for infraspecific L. polyantha (0.48%), between L. macloviana and L. occidentalis comparisons in other taxa. The Australasian species L. bris- (0.74–0.86%), and between L. polyantha and L. novae-zelandiae banica, L. novae-zelandiae, L. polyantha, and L. ruthiana also (0.78–1.2%). Lilaeopsis carolinensis and L. attenuata sequences comprise a well-supported clade. Lilaeopsis novae-zelandiae is differed by 1.2% of nucleotides. Maximum pairwise sequence not monophyletic, however, with three accessions of this spe- divergence values for all interspecific comparisons aver- cies allying with L. brisbanica, L. polyantha, and L. ruthiana in aged 2.2%. Infraspecific variation ranged from identity (for one lineage, and the remaining two accessions of the species some conspecific accessions of L. carolinensis, L. schaffneriana, comprising the other lineage. Lilaeopsis ruthiana and the clade L. novae-zelandiae, and L. occidentalis) to 1.3% of nucleotides of three accessions of L. novae-zelandiae ally most strongly; (L. novae-zelandiae ). For each species where two or more acces- the placements of L. brisbanica and L. polyantha with respect 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 797

Fig . 5. Strict consensus of 12 minimal length 541-step trees derived from maximum parsimony (MP) analysis of combined nrDNA ITS and cpDNA sequences from 37 accessions representing 12 species of Lilaeopsis and four outgroups (CI = 0.810 and 0.747, with and without uninformative characters, respectively; RI = 0.900). The majority-rule consensus tree derived from Bayesian inference analysis of these data was topologically congruent to this MP strict consensus tree. Repeating the MP analysis with 45 informative gaps scored as additional characters results in a strict consensus tree of similar topol- ogy with the exception of a monophyletic L. mauritiana arising from within a paraphyletic L. brasiliensis (inset). Numbers above nodes represent MP boot- strap and BI posterior probability values, respectively. Numbers below nodes represent Bremer support values. to the two lineages of L. novae-zelandiae vary between the ITS carolinensis was proposed by Affolter (1985) as having a South and cpDNA analyses, however. , L. caro- American origin because its North American populations linensis, L. attenuata, and L. chinensis collectively comprise a occur almost exclusively along the Gulf and Atlantic coasts, well-supported clade, with L. schaffneriana subsp. schaffneriana whereas those populations from South America are also found from Mexico and L. schaffneriana subsp. recurva from Arizona a great distance inland. The absence of inland localities in showing no to little sequence divergence. North America might reflect a relatively recent arrival, either DIVA Analysis—Two sets of DIVA analyses were per- by along-shore currents or shore-birds ( Affolter 1985 ). Both formed: the first, with terminals L. carolinensis and L. schaff- subspecies of L. schaffneriana occur in Arizona and Mexico, neriana coded with both North and South America (AB) to whereas few collections of L. schaffneriana subsp. schaffneriana reflect their current distributions; and the second, restricting have been reported from northwestern South America. Based the ancestral distribution of each of these species to only one on this limited information, the center of origin of L. schaff- of these areas (A or B), for it is possible that the other area neriana might possibly be North America. In all DIVA analy- was colonized by a recent independent dispersal. Lilaeopsis ses, the ancestral distribution of L. occidentalis was coded as 798 SYSTEMATIC BOTANY [Volume 36

Fig . 6. Bayesian inference trees derived from analyses of combined ITS and cpDNA data (including scored gaps). A. Majority-rule consensus tree, with posterior probability values indicated; B. Single tree, showing relative branch lengths.

South America in accordance with the ITS phylogenies that New Zealand was inferred, and if we assume the former sce- showed L. occidentalis (and L. masonii ) arising from within a nario, the reconstruction then suggests two consecutive dis- paraphyletic L. macloviana from South America. We presumed persals from Australia to New Zealand and three dispersals that plants of L. occidentalis, L. macloviana, and L. masonii rep- from South America to North America. If we assume the lat- resent one polymorphic species, of which the earlier name ter scenario, a dispersal from South America to New Zealand L. occidentalis would apply (see Discussion). is followed by a dispersal from New Zealand to Australia and In the first set of analyses, unrestricted optimal DIVA recon- then back again to New Zealand. structions, where maxareas = four (the total number of unit In the second set of analyses, where ancestral areas were areas inhabited by the terminals), required seven dispersal restricted to South America for L. carolinensis and to North events (not shown). For one ancestral node, however, the America for L. schaffneriana, unrestricted optimal DIVA recon- reconstruction was ambiguous with this node comprising structions required six dispersal events. For all deep ances- three different optimal distributions, including an ancestral tral nodes, however, the reconstructions were ambiguous distribution for Lilaeopsis that represented three of the areas with two nodes comprising five to nine different optimal dis- occupied by the terminals. Because none of the extant spe- tributions, including an ancestral distribution for Lilaeopsis cies of Lilaeopsis are widespread among these areas (nor is it that represented all of the areas occupied by the terminals. supposed that the ancestral taxa were either), another optimi- Restricting the number of ancestral areas assigned to each zation was carried out where the number of ancestral areas node to two also required six dispersals, but fewer alterna- assigned to each node was restricted to two. In this optimi- tive ancestral areas were inferred (Fig. 7, lower tree). In this zation, seven dispersals were still required to explain the scenario, ancestral area reconstructions for Lilaeopsis included present-day distribution of these taxa, but fewer alternative either South America or a broader region encompassing ancestral areas were inferred ( Fig. 7 , upper tree). The DIVA both North and South America. This tree resulted in more analysis resulted in a South American origin of the genus ambiguities concerning the location of dispersals between Lilaeopsis following a dispersal of its ancestor from North North and South America, and consequently, more possible America. A subsequent dispersal event to either Australia or scenarios. 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 799

L. mauritiana ) comprise a clade, but resolution of relationships is poor and weakly supported because of few parsimony informative characters. Total ITS and cpDNA sequence diver- gence between L. attenuata and L. carolinensis is 1.2%, and while this value is below the maximum average of 2.2% for other interspecific comparisons in Lilaeopsis , it is much higher than most infraspecific comparisons and, therefore, supports the specific status of each of these taxa. We maintain L. attenu- ata and L. carolinensis as separate species. Both of these species can usually be differentiated readily in the field: the former with its linear or tapering, terete leaves, and dorsal and inter- mediate mericarp ribs rounded in cross section; the latter with its spatulate (rarely linear) leaves, and dorsal and intermedi- ate ribs triangular in cross section. The presence of broadly spatulate leaves in plants of L. carolinensis and L. brasiliensis , in which a lamina has been restored by flattening and expansion of the distal portion of the rachis, has occurred twice inde- pendently during evolution of the genus. Spatulate leaves are also present in some populations of L. chinensis, but these are narrower or strap-shaped. Affolter (1985) reported that populations of L. carolinensis from the eastern U. S. A. cannot be distinguished morpho- logically from those of temperate South America. While the leaves of this amphitropic species are indeed variable, the range of variation exhibited by plants from both continents is similar. Moreover, ITS sequences for L. carolinensis acces- sions from the U. S. A., Argentina, and Portugal are identical. The plants of coastal Portugal and northwestern Spain, iden- tified by some as L. attenuata ( Almeida and Freitas 2001 ; Tutin 2001 ), are probably all L. carolinensis , as indicated previously by Affolter (1985). The lack of ITS sequence divergence among populations of L. carolinensis from three continents may be a consequence of low mutation rates and recent dispersals, or anthropogenic introductions. Petersen and Affolter (1999) described Lilaeopsis mauriti- ana for plants restricted to a single locality on the island of Mauritius. Its affinity to other species, whether those of the New World or Australasia, could not be inferred because of its distinctive fruits. Petersen and Affolter (1999) considered L. mauritiana endemic to the island, although they did not rule out completely that the species might have been recently introduced. The results of our investigations show that L. mau- Fig . 7. Results of dispersal-vicariance (DIVA) analysis for Lilaeopsis ritiana is allied closely with L. brasiliensis from South America, and outgroups. The cladograms represent simplified, fully resolved trees and that recognition of L. mauritiana as a distinct species is not of species relationships obtained from phylogenetic analyses of combined entirely clear. In trees derived from analyses of combined ITS/ ITS and cpDNA data, with each terminal representing a major lineage revealed by these analyses. The trees differ with regard to the ancestral cpDNA data, accessions of each species are commingled, and areas assigned to L. carolinensis and L. schaffneriana. Letters along inter- both MP and BI analyses of combined molecular data with nal branches indicate ancestral areas inferred with DIVA and correspond scored gaps weakly supported a monophyletic L. mauritiana to the areas indicated on the map. includes both arising from within a paraphyletic L. brasiliensis . Sequence L. macloviana and L. masonii and its ancestral distribution was coded as divergence estimates of combined data between accessions of South America; L. mauritiana was not included as a terminal (see text for further explanation). L. brasiliensis and L. mauritiana are among the lowest in inter- specific comparisons and are within the range of those values determined for L. brasiliensis only. These molecular data sug- Discussion gest that these morphologically unique plants of Mauritius Phylogenetic Resolutions— The phylogenetic results pre- might actually be aberrant members of L. brasiliensis . sented herein do not support the purported close relation- The fruit anatomy of L. brasiliensis, however, is different ship among sympatric species L. attenuata, L. brasiliensis , and from that of L. mauritiana . Spongy cells are present in all five L. carolinensis (Hill 1927; Purper and Gui Ferreira 1971; ribs of the former and absent from the latter. In L. brasiliensis Affolter 1985 ). Lilaeopsis attenuata and L. carolinensis ally as the fruit ribs are angled, whereas they are variously rounded well-supported sister taxa in trees derived from both ITS and in L. mauritiana . The complete absence of spongy cells has also combined ITS/cpDNA data, whereas in all trees L. brasiliensis been observed in other species, such as L. macloviana, L. poly- is more closely allied to L. mauritiana. In the cpDNA-derived antha , and L. ruthiana , although these species show differences trees, however, these three species (along with L. chinensis and in their rib morphologies and some populations do indeed 800 SYSTEMATIC BOTANY [Volume 36 produce spongy cells in some or all of their ribs ( Petersen and dorsal and intermediate ribs of L. occidentalis show consider- Affolter 1999 ). Affolter (1985) has reported that several spe- able variation in the extent to which they protrude above the cies of Lilaeopsis produce different fruit types in response to surface of the mericarp: low, rounded and obscure, to wing- different microhabitats, so perhaps the plants from Mauritius like and prominent. The results of the phylogenetic analyses have lost the ability to produce spongy tissue. For example, do not indicate a close relationship between L. occidentalis and the widespread South American species L. macloviana exhib- L. schaffneriana . Instead, results based on ITS sequences only its a tremendous polymorphism in fruit structure, with fruits show that accessions of L. occidentalis are inextricably linked lacking spongy cells entirely, with it confined to the lateral with those of L. macloviana , a species distributed widely in ribs only, or with abundant spongy cells throughout all five western South America, from Colombia south to Tierra del ribs. The fruit ribs may be low and obscure, to prominent Fuego, and eastward to the Falkland Islands. Results of the and broadly rounded or projecting. Populations of L. brasil- combined ITS/cpDNA data revealed a sister group relation- iensis are known from river banks, ditches, seepage areas, ship between L. occidentalis and L. macloviana , albeit only a and marshes or bogs ( Affolter 1985 ), whereas L. mauriti- single accession of L. macloviana was included. ana occurs in or along a moderately flowing, clear-watered Lilaeopsis macloviana occurs in a large variety of habitats, stream ( Petersen and Affolter 1999 ). Lilaeopsis brasiliensis has from sea level in the brackish water of intertidal zones, along never been known to produce fruits lacking spongy tissue, margins of ditches and lakes, to wet meadows and bogs up to but it may be possible that the different habitat of Mauritius 4,700 m elevation ( Affolter 1985 ). In the northern portion of has induced this response. The rapid loss of dispersal abilities its range, the species extends into the tropics along the Andes. of these plants on Mauritius may also explain the differences It is polymorphic, showing considerable variation in vegeta- in fruit structure between these two species. Whether L. mau- tive organs and fruit morphology and anatomy, and prior to ritiana should be maintained as a distinct species, or treated Affolter’s monograph it was treated as six species and sev- as some infraspecific rank of L. brasiliensis, or even subsumed eral infraspecific taxa. The conspicuous variation in vegeta- under L. brasiliensis , remains the subject of further sampling tive morphology exhibited by these plants is accounted for by of these taxa from both South America and Mauritius. phenotypic plasticity in response to the variety of microenvi- Lilaeopsis chinensis (eastern grasswort), a state-listed rare ronments in which they grow ( Affolter 1985 ). A continuous species, is found along the Atlantic and Gulf coasts of North series of fruit types is apparent according to the distribu- America ( Affolter 1985 ; Moore et al. 2009 ). The three acces- tion of spongy cells, ranging from abundant in all five ribs sions included in this study, all from the southeastern U. S. A., to none whatsoever, and several different fruit types may be comprise a monophyletic group. In North America, the spe- present in a single collection. The shape and the size of the cies is sympatric with L. carolinensis and they share several fruits, as well as the prominence of dorsal and intermediate vegetative traits, including a flattening of the distal portion of ribs and the number of vittae, are also variable within indi- the leaves. Indeed, in the absence of mature fruits, these spe- vidual collections. Our sampling of this polymorphic species cies may be confused when their leaves are of similar shape is sparse, with only four accessions included in the ITS study; and size, yet Hill (1927) and Affolter (1985) found no evi- however, they represent collections from throughout its range dence that these two taxa should be considered a single spe- (i.e. Argentina, Bolivia, Peru, and Falkland Islands). Lilaeopsis cies. Our results support the segregation of L. chinensis from occidentalis exhibits similar variability as L. macloviana in the L. carolinensis . degree to which dorsal and intermediate ribs protrude above Lilaeopsis schaffneriana is distributed in southeastern Arizona the surface of the mericarp. Its fruits have spongy tissue con- and adjacent Sonoran Mexico, northern and central Mexico, fined only to the two lateral ribs, a condition also occurring and northwestern South America. A single, sterile collection in L. macloviana . Both species have leaves borne in clusters is known from Santo Domingo, Dominican Republic and was at the apex of vertical rhizome branches (although they may only tentatively identified as this species because of its veg- also occur individually and along horizontal rhizomes in etative similarity and close proximity to L. schaffneriana from L. occidentalis). Overall, the two species are similar morpho- Mexico (Affolter 1985). Two subspecies are currently recog- logically, and the range of variation exhibited by L. occidentalis nized, with subsp. recurva (Huachuca water umbel) being can largely be accommodated within that variability seen in federally endangered (Titus and Titus 2008). Differences L. macloviana. The close relationship between these two spe- between the two subspecies include the overall shape of cies is further supported by a shared chromosome number their desiccated, mature fruits and their geographic separa- of n = 22, a polyploid number also known only for L. masonii tion. All examined accessions of L. schaffneriana comprise a ( Affolter 1985 ). All other species of Lilaeopsis for which cyto- monophyletic group in the phylogenetic analyses, with no to logical data are available have a base chromosome number little sequence divergence among them. Lilaeopsis schaffneri- of x = 11 ( Affolter 1985 ), a cytotype that is prevalent within ana is distantly related to L. macloviana from western South subfamily Apioideae ( Moore 1971 ). The habitat of L. occi- America, a sympatric species in the northern portion of its dentalis (brackish water of intertidal zones) also falls within range where some populations are almost indistinguishable that favored by plants of L. macloviana growing at sea level. from those of L. schaffneriana . Pairwise nucleotide divergence estimates between acces- Lilaeopsis occidentalis is distributed along the Pacific coast of sions of L. occidentalis and L. macloviana were generally low North America, from the Queen Charlotte Islands southward and these accessions are mixed in the ITS trees. In addition to Marin County, California, and is almost entirely confined to the molecular and phylogenetic results presented herein, to salt water or brackish habitats, frequently below the level morphological, fruit anatomical, cytological, and ecological of high tides (Affolter 1985). The absence of spongy cells from data all suggest that these plants probably best represent one the dorsal and intermediate fruit ribs of this species is a useful polymorphic species, of which the earlier name L. occidenta- character separating it from the vegetatively similar L. schaff- lis would apply. This species has an amphitropic distribution neriana, where spongy cells are present in all five ribs. The in the western hemisphere, with extensions into the tropics 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 801 along the Andes. Such a widespread geographic range is also ana, collectively form a well-supported monophyletic group shown by L. carolinensis , an Atlantic coastal species. in trees resulting from both ITS and combined ITS/cpDNA Lilaeopsis masonii (Mason’s lilaeopsis) is a California state- data. In trees derived from cpDNA sequences only, with or listed rare species that has recently been subsumed under without scored gaps, these Australasian taxa comprise two L. occi dentalis (Fiedler et al. in press). Traditionally, L. masonii separate clades, a result of too few phylogenetically informa- was considered restricted almost exclusively to the Sacra- tive characters. In the combined trees, the five accessions of mento/San Joaquin River Delta region of California and was L. novae-zelandiae separated into two lineages, with L. bris- distinguished from the more widespread L. occidentalis by banica, L. polyantha, and L. ruthiana allying with one of them. its narrower, usually shorter leaves with few, obscure septa Within this Australasian clade, sequence divergence esti- (Mathias and Constance 1977; Affolter 1985). In contrast, mates for combined ITS/cpDNA data ranged from identity to plants of L. occidentalis were described as more robust and dis- 1.3%, with the greatest differences found between the two lin- tributed along the Pacific coast, although inland populations eages of L. novae-zelandiae; otherwise, divergence estimates for with smaller leaves were noted (Affolter 1985). Subsequent some interspecific comparisons were low (e.g. 0.09% between field surveys by Fiedler and colleagues expanded the range of L. ruthiana and L. novae-zelandiae ), falling into the range for L. masonii and showed that the leaf vegetative characters tradi- infraspecific comparisons in other taxa. The relationships tionally used to separate it from L. occidentalis overlap consid- among these Australasian taxa are not easily resolved. One erably (reviewed in Fiedler et al. in press). Both species have could suppose either a monophyletic L. novae-zelandiae to sub- similar vegetative and fruit morphologies, are sympatric and sume L. brisbanica, L. polyantha, and L. ruthiana , or maintain grow in similar brackish habitats, and share the same unusual each of these Australasian species as distinct and recognize polyploid chromosome number. Additional evidence provided L. novae-zelandiae accessions 2674 and 2677 as a new species. by Fiedler et al. (in press) for recognition of these two species The latter accessions, however, are not particularly distinct, as one taxon included essentially identical ITS sequences and either morphologically or geographically, and accession no. a high degree of genetic similarity, as discerned through AFLP 2677 (Affolter 182 ) was collected from the type locality of the analyses. We agree with Fiedler et al. (in press) that the rare species. The vegetative similarities and overlapping patterns L. masonii should no longer be recognized as a separate taxon of fruit variation observed among these Australian taxa, cou- and that it, like L. macloviana , be subsumed within L. occidentalis . pled with low sequence divergence estimates observed in Affolter (1985) consolidated the three New Zealand spe- some interspecific comparisons and the paraphyly of L. novae- cies of Lilaeopsis recognized by Hill (1927, 1928 , 1929 ) into the zelandiae , suggest that the entire group be treated as a sin- single polymorphic taxon L. novae-zelandiae. The fruit of this gle, polymorphic species. However, before such a treatment species shows continuous variation in shape, size, distribu- is invoked, further molecular and morphological studies of tion and presence of spongy tissue, and shape of the ribs. In these Australasian species are warranted, given the limited addition, Affolter described L. ruthiana from New Zealand, sampling herein of such a taxonomically complex group. which differs from L. novae-zelandiae by its relative uniformity Biogeography— Because a great proportion of the distribu- in fruit structure. In L. ruthiana the mericarps are generally tional range of Lilaeopsis parallels the Gondwanan disjunction semicircular in cross section, the ribs are low and rounded, in the southern hemisphere, the genus was assumed to have and spongy cells are absent from all five ribs or scarcely a southern hemispheric and ancient origin (Raynal 1977). present in the lateral ribs only. Some fruits of L. novae-zelan- However, biogeographic studies of other angiosperm genera diae, however, have a similar external appearance to that of exhibiting amphitropic distribution patterns with transoceanic L. ruthiana. Furthermore, Clayton (1998) has observed that disjunctions in the southern hemisphere have revealed that L. ruthiana is almost indistinguishable from L. novae-zelandiae the predominant direction of dispersal was from the northern and suggested that its taxonomic distinction may not be justi- to the southern hemisphere, even for those groups that are fied. Affolter also merged the three Australian species recog- much diversified in the latter (Raven 1963); moreover, such nized by Hill into another polymorphic species, L. polyantha . long-distance dispersals were likely relatively recent events The Australian species L. brisbanica was erected for specimens (reviewed in Spalik et al. 2010). Genera of Apiaceae inferred previously ascribed to both L. novae-zelandiae and L. polyan- to have ancestral distributions in the northern hemisphere tha ( Affolter 1985 ; Bean 1997 ). In describing L. brisbanica , Bean with subsequent recent, long-distance dispersals to the south- (1997) compared these specimens to only the typical form of ern hemisphere include Sanicula L. (Vargas et al. 1998, 1999 ), L. polyantha and not the broadened circumscription of L. poly- Osmorhiza Raf. ( Wen et al. 2002 ; Yoo et al. 2002 ), Chaerophyllum antha, as recognized by Affolter. As such, the diagnostic fea- L./Oreomyrrhis Endl. (Chung et al. 2005), Daucus L. ( Spalik tures of L. brisbanica fall within the range of variation seen in and Downie 2007 ; Spalik et al. 2010 ), Eryngium L. ( Calviño L. polyantha sensu Affolter. There are few, if any, vegetative et al. 2008), and Apium L. (Spalik et al. 2010). Other Apiaceae characters useful for distinguishing among L. novae-zelandiae, genera showing American amphitropic disjunctions include L. polyantha , and L. ruthiana and they are often impossible to Ammoselinum Torr. & A. Gray and Spermolepis Raf. ( Constance tell apart in the absence of fruits (Affolter 1985). As in other 1963), but their ancestral areas and directions of dispersal/ species of Lilaeopsis, vegetative features display considerable migration have yet to be determined through modern biogeo- variation in response to the microhabitat in which they are graphic reconstructions. Prior to this study, we were unaware found. Lilaeopsis novae-zelandiae and L. polyantha show over- of any umbellifer genus exhibiting an American amphi- lapping variation among fruit characters, with fruits of some tropic distribution pattern with transoceanic disjunctions in populations of the former impossible to key apart from some the southern hemisphere that may have had its origin in the populations of the latter; thus, these two species are best cir- southern hemisphere. cumscribed geographically ( Affolter 1985 ). Of the DIVA analyses presented, we favor the tree assum- The four Australasian species of Lilaeopsis included in this ing terminal polymorphisms (AB) for both L. carolinensis and study, L. brisbanica, L. novae-zelandiae, L. polyantha , and L. ruthi- L. schaffneriana ( Fig. 7 , upper tree), in accord with their present 802 SYSTEMATIC BOTANY [Volume 36 distributions, rather than inferring a single ancestral region capable of overland migration along the Cordilleran system. for each of these lineages that may be regarded as specula- Lilaeopsis schaffneriana occurs in the northern Andes, central tive (Fig. 7, lower tree). The optimal solution of DIVA sug- and northern Mexico, and adjacent Arizona, and may even- gests that South America is indeed the ancestral area for tually be found to be fairly continuously distributed through L. carolinensis, with a dispersal to North America occurring the montane neo-tropics. relatively recently at a terminal branch. A similar scenario Lilaeopsis is well represented in cool temperate regions of may be invoked for L. schaffneriana. The results of DIVA indi- the southern hemisphere, with 10 species occurring in South cate that Lilaeopsis probably originated in South America, America and Australasia (Australia, Tasmania, and New following a dispersal of its ancestor from North America. Zealand). Additional populations are known from Mauritius, A subsequent dispersal from South America to Australia (or Madagascar, and the Kerguelen Archipelago, and no doubt New Zealand), dispersals from Australia to New Zealand these inconspicuous plants will eventually be reported (or between New Zealand and Australia), and three disper- from other scattered islands of the South Atlantic and South sal events from South America to North America were also Indian Oceans. The presence of Lilaeopsis in Mauritius and inferred. Madagascar, a species representing either a local endemic A minimum of seven dispersal events is required to explain (L. mauritiana) or an aberrant form of L. brasiliensis, is likely a the present-day distribution of Lilaeopsis , and the isolated recent introduction. In most phylogenies inferred herein, the and scattered presence of Lilaeopsis in coastal Portugal and accessions of L. mauritiana and L. brasiliensis are commingled, Spain, Mauritius, Madagascar, the Dominican Republic, and implying a recent split of the Mauritius taxon. One optimal the Kerguelen Archipelago are recent introductions of pos- solution of DIVA indicates that Lilaeopsis was dispersed from sibly anthropogenic origins. Biogeographic studies, includ- South America to New Zealand, with a subsequent disper- ing those of several genera of Apiaceae, have suggested that sal westward from New Zealand to Australia and then back bird-mediated oceanic dispersal underlies the distribution to New Zealand again. Dawson (1971) hypothesized that of many trans-Pacific disjunct plant groups, especially those Lilaeopsis and other Apiaceae that occur in wet habitats in favoring wetland and aquatic habitats (Vargas et al. 1998, New Zealand arrived there by overseas, long-distance dis- 1999 ; Winkworth et al. 2002; Les et al. 2003; Sanmartín and persal, probably a result of transport by migratory birds, a Ronquist 2004 ; Chung et al. 2005 ; de Queiroz 2005 ). Similarly, scenario reported repeatedly to explain many other south- amphitropic disjunctions in North and South America have ern hemisphere plant disjunctions (Sanmartín and Ronquist also been explained by long-distance dispersal via migrating 2004 ; de Queiroz 2005 ). Winkworth et al. (2002) have postu- birds ( Raven 1963 ; Cruden 1966 ). Fruits of Lilaeopsis can retain lated that many contemporary alpine species of New Zealand their buoyancy in both fresh and salt water for many months, are recent arrivals, reaching New Zealand or diversifying with only a minimal loss of seed viability, and their dispersal there in the late Tertiary, and then traveling to other southern by sea currents or birds may have facilitated their transport hemispheric landmasses. In contrast, others have reported to new regions (Affolter 1985). The optimal solution of DIVA that many alpine plant groups entered New Zealand via indicates multiple dispersals of Lilaeopsis between North and Australia, following the direction of the prevailing wester- South America, and while such intercontinental disjunctions lies (e.g. Raven 1973 ; Pole 1994 ). Another optimal solution can be simply explained by long-distance dispersal by migra- of DIVA indicates that Lilaeopsis was dispersed from South tory birds, a more or less continuous overland migration America to Australia, with subsequent dispersals eastward to between continents can also be invoked. Affolter (1985) indi- New Zealand. Further resolution of biogeographic relation- cated two avenues of migration of Lilaeopsis between North ships will come from increased study of these Australasian and South America: one on the eastern sides of the conti- plants. nents, the other between the northern Andes and southwest- Cladistic analysis of molecular data from both the chloro- ern North America. “Island hopping” across exposed regions plast and nuclear genomes of Lilaeopsis has revealed infrage- of the continental shelf along the Gulf and Atlantic coasts dur- neric relationships heretofore unknown and demonstrated ing episodes of sea level lowering during the Pleistocene gla- that several previously recognized species inadequately ciations afforded opportunities for plant migration, with their reflect natural groupings. The genus has probably had a intervening range wiped out with the final rise in sea level South American origin following a dispersal of its ancestor ( Flint 1971 ; Affolter 1985 ). “Mountain hopping” through the from North America. In other angiosperms having a simi- American Cordillera was an important migratory route for lar distribution pattern, the predominant direction of dis- temperate species moving across the American tropics, and persal appears to have been the opposite, from the northern such a pathway would have become available first during the to southern hemisphere. Further studies of the genus are in Miocene and through the Pliocene ( Constance 1963 ; Raven order, especially to confirm the taxonomic changes proposed 1963 ; Cruden 1966 ). Raven (1963) hypothesized that American herein. However, because of their inconspicuous nature and amphitropic plant disjunctions would have occurred from the grass-like appearance in sterile condition, these plants are Pliocene onwards, as neither species nor the habitats to which often overlooked in the field and not well represented in her- they are adapted probably existed before this time. This time- baria, thus a substantial effort will be required. The extent of frame is in accordance with molecular dating for several variability in fruit characters exhibited by several species also apioid umbellifer genera and other taxa exhibiting amphi- needs to be reexamined to produce a classification that is both tropic distribution patterns ( Spalik et al. 2010 ). During the functional and reflects the phylogenetic history of the plants. Pleistocene, wetland habitats from northern South America Three taxa have yet to be considered in molecular study to central Mexico were extensive, maximizing the exchange of (L. attenuata subsp. ulei, L. fistulosa, and L. tenuis), and their Lilaeopsis species between North and South America ( Affolter inclusion in future analyses should reveal their phylogenetic 1985 ). The presence of L. macloviana at elevations up to 4,700 placements and taxonomic status, as well as confirm or revise m in the Andes indicates that these plants are physiologically the biogeographic scenarios inferred herein. 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 803

Acknowledgments. We thank the curators of herbaria cited in the J. Lahham , and A. El-Oqlah . 2001 . Tribes and clades within Apiaceae text for kindly providing herbarium specimens, many of which pro- subfamily Apioideae: the contribution of molecular data . Edinburgh vided DNA for use in the molecular studies; G. Petersen, University of Journal of Botany 58: 301 – 330 . Copenhagen, for providing plant material and DNAs for several spe- Farris, J. S., M. Källersjö , A. G. Kluge , and C. Bult . 1995 . Testing signifi- cies; Florida Aquatic Nurseries (Florida, U. S. A.) and Tropica Aquarium cance of incongruence . Cladistics 10: 315 – 319 . Plants (Denmark) for supplying living plants; P. L. Fiedler, E. K. Crumb, Feist, M. A. E. and S. R. Downie . 2008 . A phylogenetic study of Oxypolis and A. K. Knox for access to unpublished ITS sequences and manuscript; and Ptilimnium (Apiaceae) based on nuclear rDNA ITS sequences. C. Calviño, J. Cordes, C. Danderson, M. A. Feist, V. Sagun, H. Hines, Systematic Botany 33: 447 – 458 . C. Rasmussen, K. Paige, A. Edwards, and C. Shearer for advice during Fiedler, P. L., E. K. Crumb , and A. K. Knox . 2011 . Reconsideration of the the course of this study; and C. Calviño and anonymous reviewers for taxonomic status of Mason’s Lilaeopsis – a state-protected rare species comments on earlier versions of the manuscript. This paper represents, in California . Madroño (in press). in part, a M. S. thesis submitted by T. Bone to the Graduate College of Flint, R. F. 1971 . Glacial and Quaternary geology . New York : John Wiley and the University of Illinois at Urbana-Champaign in 2007. This work was Sons . supported by funding from the Francis M. and Harley M. Clark Research Hardway, T. M., K. Spalik , M. F. Watson , D. S. Katz-Downie , and S. R. Support Fund, the John R. Laughnan Travel Fund, a travel grant from the Downie . 2004 . Circumscription of Apiaceae tribe Oenantheae . South Graduate College of the University of Illinois to T. Bone, and by NSF grant African Journal of Botany 70: 393 – 406 . DEB 0089452 to S. Downie. Hill, A. W . 1927 . The genus Lilaeopsis : a study in geographical distribution . Journal of the Linnean Botanical Society of London 47: 525 – 551 . Hill, A. W . 1928 . A new species of Lilaeopsis from New Zealand . Kew Literature Cited Bulletin 1928: 266 – 267 . Hill, A . W. 1929. Lilaeopsis in Tasmania and New Zealand . Kew Bulletin Affolter, J. M . 1985 . A monograph of the genus Lilaeopsis (Umbelliferae) . 1929 : 119 – 121 . Systematic Botany Monographs 6: 1 – 140 . Jeanmougin, F., J. D. Thompson , M. Gouy , D. G. Higgins , and T. J. Gibson . Almeida, J. D. and H. Freitas . 2001 . The exotic and invasive flora of 1998 . Multiple sequence alignment with CLUSTAL X . Trends in Portugal . Botanica Complutensis 25: 317 – 327 . Biochemical Sciences 23: 403 – 405 . Barker, K. F. and F. M. Lutzoni . 2002 . The utility of the incongruence length Kaplan, D. R . 1970 . Comparative development and morphological inter- difference test . Systematic Biology 51: 625 – 637 . pretation of ‘rachis-leaves’ in Umbelliferae. Botanical Journal of the Bean, A. R. 1997 . Lilaeopsis brisbanica (Apiaceae), a new species from Linnean Society 63 (Suppl. 1) : 101 – 125 . Queensland, Australia. Austrobaileya 5: 145 – 148 . Lee, C.-S. and S. R. Downie . 2006 . Phylogenetic relationships within Cicuta Bone, T. S. 2007 . A phylogenetic and biogeographic study of the umbellifer genus (Apiaceae tribe Oenantheae) inferred from nuclear rDNA ITS and Lilaeopsis. M. S. thesis. Urbana-Champaign : University of Illinois . cpDNA sequence data . Canadian Journal of Botany 84: 453 – 468 . Bremer, K . 1994 . Branch support and tree stability . Cladistics 10: 295 – 304 . Les, D. H., D. J. Crawford , R. T. Kimball , M. L. Moody , and E. Landolt . Briquet, J . 1897 . Recherches sur les feuilles septées chez les Dicotylédones . 2003 . Biogeography of discontinuously distributed hydrophytes: a Bulletin of the Boissier Herbarium 5: 453 – 468 . molecular appraisal of intercontinental disjunctions . International Calviño, C. I., S. G. Martínez , and S. R. Downie . 2008 . The evolutionary Journal of Plant Sciences 164: 917 – 932 . history of Eryngium (Apiaceae, Saniculoideae): rapid radiations, long Mathias, M. E. and L. Constance . 1977 . Two new local Umbelliferae distance dispersals, and hybridizations . Molecular Phylogenetics and (Apiaceae) from California . Madroño 24: 78 – 83 . Evolution 46: 1129 – 1150 . Moore, D. M. 1971 . Chromosome studies in Umbelliferae . Pp. 233 – 255 in Charlton, W. A. 1992 . The rachis-leaves of Lilaeopsis brasiliensis (Glaziou) J. M. The biology and chemistry of the Umbelliferae , ed. V. H. Heywood . New Affolter (Apiaceae). Botanical Journal of the Linnean Society 109: 259 – 280 . York : Academic Press . Chung, K.-F., C.-I. Peng , S. R. Downie , K. Spalik , and B. A. Schaal . 2005 . Moore, G. E., C. R. Peter , D. M. Burdick , and D. R. Keirstead . 2009 . Status Molecular systematics of the trans-Pacific alpine genus Oreomyrrhis of the eastern grasswort, Lilaeopsis chinensis (Apiaceae), in the Great (Apiaceae): phylogenetic affinities and biogeographic implications. Bay Estuary, New Hampshire, U. S. A . Rhodora 111: 171 – 188 . American Journal of Botany 92: 2054 – 2071 . Nylander, J. A. A . 2004 . MrModeltest v2. Evolutionary Biology Centre , Clayton, J. 1998 . New Zealand native plants for use in aquariums and Uppsala University : Distributed by the author . ponds . Website: www.thekrib.com/Plants/Plants/NZ/ (accessed Petersen, G. and J. Affolter . 1999 . A new species of Lilaeopsis (Apiaceae) October, 2009 ). from Mauritius . Novon 9: 92 – 94 . Constance, L . 1963 . Amphitropical relationships in the herbaceous flora Petersen, G., O. Seberg , and S. Larsen . 2002 . The phylogenetic and taxo- of the Pacific Coast of North and South America: a symposium. nomic position of Lilaeopsis (Apiaceae), with notes on the applicabil- Introduction and historical review . The Quarterly Review of Biology 38: ity of ITS sequence data for phylogenetic reconstruction . Australian 109 – 116 . Systematic Botany 15: 181 – 191 . Cruden, R. W . 1966 . Birds as agents of long-distance dispersal for disjunct Pole, M . 1994 . The New Zealand flora—entirely long-distance dispersal? plant groups of the temperate Western Hemisphere . Evolution 20: Journal of Biogeography 21: 625 – 635 . 517 – 532 . Purper, C. and A. Gui Ferreira . 1971 . Pollen grains of Umbelliferae from Dawson, J. W. 1971 . Relationships of the New Zealand Umbelliferae . Rio Grande do Sul. II. Genera Bowlesia and Lilaeopsis . Review of Pp. 43 – 61 in The biology and chemistry of the Umbelliferae , ed. V. H. Brasilian Biology 31: 103 – 106 . Heywood . New York : Academic Press . Raven, P. H . 1963 . Amphitropical relationships in the floras of North and de Queiroz, A . 2005 . The resurrection of oceanic dispersal in historical bio- South America. The Quarterly Review of Biology 38: 151 – 177 . geography . Trends in Ecology & Evolution 20: 68 – 73 . Raven, P. H . 1973 . Evolution of subalpine and alpine plant groups in New Downie, S. R. and D. S. Katz-Downie . 1996 . A molecular phylogeny of Zealand . New Zealand Journal of Botany 11: 177 – 200 . Apiaceae subfamily Apioideae: evidence from nuclear ribosomal Raynal, J . 1977 . Le genre Lilaeopsis (Ombellifères) à Madagascar. Adansonia DNA internal transcribed spacer sequences. American Journal of 17: 151 – 154 . Botany 83: 234 – 251 . Ronquist, F. 1996 . DIVA version 1.1 computer program and manual avail- Downie, S. R. and D. S. Katz-Downie . 1999 . Phylogenetic analysis of chlo- able from anonymous FTP from Uppsala University (ftp.uu.se or ftp. roplast rps16 intron sequences reveals relationships within the woody systbot.uu.se). southern African Apiaceae subfamily Apioideae . Canadian Journal of Ronquist, F . 1997 . Dispersal-vicariance analysis: a new approach to the quan- Botany 77: 1120 – 1135 . tification of historical biogeography. Systematic Biology 46: 195 – 203 . Downie, S. R., D. S. Katz-Downie , F.-J. Sun , and C.-S. Lee . 2008 . Phylogeny Ronquist, F. and J. P. Huelsenbeck . 2003 . MrBayes3: Bayesian phylogenetic and biogeography of Apiaceae tribe Oenantheae inferred from nuclear inference under mixed models . Bioinformatics 19: 1572 – 1574 . rDNA ITS and cpDNA psbI -5′trnK(UUU) sequences, with emphasis on Sales, F., I. Hedge , A. X. P. Coutinho , and A. Marques . 2004 . Apiaceae sub- the North American Endemics clade . Botany 86: 1039 – 1064 . family Apioideae in Madagascar . South African Journal of Botany 70: Downie, S. R., D. S. Katz-Downie , and M. F. Watson . 2000 . A phylogeny of 446 – 448 . the family Apiaceae based on chloroplast DNA rpl16 Sanmartín, I. and F. Ronquist . 2004 . Southern hemisphere biogeogra- and rpoC1 intron sequences: towards a suprageneric classification of phy inferred by event-based models: plant versus animal patterns . subfamily Apioideae. American Journal of Botany 87: 273 – 292 . Systematic Biology 53: 216 – 243 . Downie, S. R., G. M. Plunkett , M. F. Watson , K. Spalik , D. S. Katz-Downie , Simmons, M. P. and H. Ochoterena . 2000 . Gaps as characters in sequence- C. M. Valiejo-Roman , E. I. Terentieva , A. V. Troitsky , B.-Y. Lee , based phylogenetic analyses . Systematic Biology 49: 369 – 381 . 804 SYSTEMATIC BOTANY [Volume 36

Sorenson, M. D. and E. A. Franzosa . 2007 . TreeRot, version 3. Boston: California Botanical Garden, Constance C-1227 (MO 2260055), DNA acces- Boston University . Website: people.bu.edu/msoren/TreeRot.html sion no. 2920 (ITS: GU128110). U. S. A., Louisiana, St. John Parish, rare on (accessed December, 2010). stumps in canal along Hwy 51, 12 April 1992, Montz & Montz 5696 (LSU), Spalik, K. and S. R. Downie . 2006 . The evolutionary history of Sium sensu DNA accession no. 3097 (ITS: GU128111). U. S. A., Louisiana, St. Charles lato (Apiaceae): dispersal, vicariance, and domestication as inferred Parish, New Orleans District, Bonnet Carre Spillway, 25 April 1981, Montz from ITS rDNA phylogeny . American Journal of Botany 93: 747 – 761 . 5191 (LSU), DNA accession no. 3098 (ITS: GU128112). U. S. A., Louisiana, Spalik, K. and S. R. Downie . 2007 . Intercontinental disjunctions in St. Charles Parish, uncommon in slip along Burchell Canal, Bayou Gauche, Cryptotaenia (Apiaceae, Oenantheae): an appraisal using molecular 20 July 1994, Montz 6836 (LSU), DNA accession no. 3099 (cpDNA: data . Journal of Biogeography 34: 2039 – 2054 . GU144642, GU144616; ITS: GU128113). U. S. A., Louisiana, East Baton Spalik, K., M. Piwczyn´ski , C. A. Danderson , R. Kurzyna-Młynik , T. S. Rouge Parish, growing just beneath and on the surface of the Capitol Bone , and S. R. Downie . 2010 . Amphitropic amphiantarctic disjunc- Lakes in the shallows near the bank, 21 March 1995, Walker 6 (LSU), DNA tions in Apiaceae subfamily Apioideae . Journal of Biogeography 37: accession no. 3103 (ITS: GU128114). U. S. A., cultivated, origin unknown, 1977 – 1994 . Bogner s. n., 1985, DNA and leaf material supplied by G. Petersen, Petersen Swofford, D. L. 2002 . PAUP* Phylogenetic analysis using parsimony (*and GPL4 (C, ILL), DNA accession no. 2720 (cpDNA: EF185225; ITS: AF466276). other methods), v. 4.0 beta 10. Sunderland: Sinauer Associates. U. S. A., North Carolina, New Hanover Co., Wilmington, 31 May 1987, Titus, P. J. and J. H. Titus . 2008 . Ecological monitoring of the endan- MacDougal 2071 (MO 05033978), DNA accession no. 2399 (cpDNA: gered Huachuca water umbel (Lilaeopsis schaffneriana subsp. recurva : GU144643, GU144617; ITS: GU128115). Origin unknown, cultivated mate- Apiaceae) . The Southwestern Naturalist 53: 458 – 465 . rial identified as L. macloviana obtained from Tropica Aquarium Plants, Tutin, T. G. 2001 . Umbelliferae . Pp. 315 – 375 in Flora Europaea , vol. 2 . eds. Denmark, Troels 040D , Bone 109 (ILL), DNA accession no. 3180 (cpDNA: T. G. Tutin , V. H. Heywood , N. A. Burges , D. M. Moore , D. H. GU144644, GU144618; ITS: GU128116). L. chinensis (L.) Kuntze: U. S. A., Valentine , S. M. Walters and D. A. Webb . Cambridge : Cambridge Louisiana, St. Marys Parish, along edge of Wax Bayou, April 1989, Givens University Press . 5167 (LSU), DNA accession no. 3101 (ITS: GU128117). U. S. A., Louisiana, Vargas, P., B. G. Baldwin , and L. Constance . 1998 . Nuclear ribosomal DNA St. Tammany Parish, Pearl River Wildlife Management Area south of Hwy evidence for a western North American origin of Hawaiian and South 90, marshy environment, 20 April 1991, Kantz 22 (LSU), DNA accession American species of Sanicula (Apiaceae) . Proceedings of the National no. 3100 (cpDNA: GU144645, GU144619; ITS: GU128118). U. S. A., North Academy of Sciences USA 95: 235 – 240 . Carolina, New Hanover Co., west bank of Cape Fear River, 31 May 1987, Vargas, P., B. G. Baldwin , and L. Constance . 1999 . A phylogenetic study MacDougal 2068 (MO 05033977), DNA accession no. 2401 (cpDNA: of Sanicula sect. Sanicoria and S . sect. Sandwicenses (Apiaceae) based GU144608; ITS: EF177714). L. macloviana (Gand.) A. W. Hill: Argentina, on nuclear rDNA and morphological data . Systematic Botany 24: Tierra del Fuego, Estancia Cullen, spring along road to Puesto Beta, 228 – 248 . 7 January 1971, Goodall 3234 (UC 150110), DNA accession no. 2921 (ITS: Wen, J. and E. A. Zimmer . 1996 . Phylogeny and biogeography of Panax L. (the GU128119). Bolivia, Dep. Oruro, Prov. L. Cabrera, 28 February 1986, Beck ginseng genus, Araliaceae): inferences from ITS sequences of nuclear 11830 (UC 1549364), DNA accession no. 2925 (cpDNA: GU144609; ITS: ribosomal DNA . Molecular Phylogenetics and Evolution 6: 167 – 177 . GU128120). Falkland Islands, East Falkland, San Carlos, White Rincon, Wen, J., P. P. Lowry II , J. L. Walck , and K.-O. Yoo . 2002 . Phylogenetic and moist sand at top of beach, 24 January 1964, Moore 649 (UC 297946), DNA biogeographic diversification in Osmorhiza (Apiaceae) . Annals of the accession no. 2923 (ITS: GU128121). Peru, Cuzco, 15 km S of Cuzco on Missouri Botanical Garden 89: 414 – 428 . road to Urcos, cult. University of Michigan Botanical Gardens, Affolter 119 Winkworth, R. C., S. J. Wagstaff , D. Glenny , and P. J. Lockhart . 2002 . Plant (MICH, GA), DNA accession no. 2518 (ITS: EF177715). L. masonii Mathias dispersal N.E.W.S from New Zealand. Trends in Ecology & Evolution & Constance: U. S. A., California, Sacramento Co., Twitchell Island, 24 17: 514 – 520 . August 2007, Fiedler et al. s. n. (SFSU) (ITS: HQ634700). U. S. A., California, Yoder, A. D., J. A. Irwin , and B. A. Payseur . 2001 . Failure of the ILD to Solano Co., Liberty Island, 23 August 2007, Fiedler et al. s. n. (SFSU) (ITS: determine data combinability for slow loris phylogeny. Systematic HQ647238). L. mauritiana G. Petersen & Affolter: Mauritius, Le Val Biology 50: 408 – 424 . Nature Park, Windeløv s. n., 3 May 1992, DNA and leaf material supplied Yoo, K.-O., P. P. Lowry II , and J. Wen . 2002 . Discordance of chloroplast by G. Petersen, Petersen GPL8 (C, ILL), DNA accession no. 2151 (cpDNA: and nuclear ribosomal DNA data in Osmorhiza (Apiaceae) . American EF185226; ITS: AF466277). Mauritius, Le Val Nature Park, Affolter 000 Journal of Botany 89: 966 – 971 . (GA), DNA accession no. 2654 (cpDNA: GU144646, GU144620; ITS: GU128122). Mauritius, cultivated material obtained from Tropica Appendix 1. List of Lilaeopsis and outgroup specimens examined, Aquarium Plants, Denmark, Troels 040B, Bone 106 (ILL), DNA accession with source and voucher information and GenBank accession numbers. no. 3181 (cpDNA: GU144647, GU144621; ITS: GU128123). L. novae- Two GenBank numbers for cpDNA indicate separate sequences for the zelandiae (Gand.) A. W. Hill: New Zealand, North Island, Taranaki, nr rps16 intron and rps16 -trnK intergenic spacer regions, respectively; a sin- mouth of Kaupokonui Strm, coast WNW of Manaia, Affolter 168 (MICH, gle GenBank number for cpDNA indicates contiguous sequence from the GA), DNA accession no. 2674 (cpDNA: GU144648, GU144622; ITS: rps16 intron to trnK (i.e., including the rps16 3′exon). GU128124). New Zealand, South Island, Otago, Otago Peninsula, Tomahawk Lagoon, Affolter 182 (MICH, GA), DNA accession no. 2677 Lilaeopsis attenuata (Hook. & Arn.) Fern. subsp. attenuata : Argentina, (cpDNA: GU144649, GU144623; ITS: GU128125). New Zealand, origin Corrientes, Dep. Mburucuyá, Estancia Santa Teresa, cult. University of unknown, cultivated, DNA and leaf material supplied by G. Petersen, Michigan Botanical Gardens, Affolter 115 (MICH, GA), DNA accession no. Petersen GPL9 (C, ILL), DNA accession no. 2722 (cpDNA: EF185227, ITS: 2666 (cpDNA: GU144637, GU144611; ITS: EF177710). L. brasiliensis AF466278). New Zealand, North Island, Auckland Ecological Region, (Glaz.) Affolter: Argentina, Corrientes, Sloping ground towards a swamp, Tamaki Ecological District, east bank of the Tamaki River, 5 November 11 June 1980, Crucecita s. n. (UC 1538838), DNA accession no. 2915 (cpDNA: 2000, Gardner 10225 (AK), DNA accession no. 3177 (cpDNA: GU144650, GU144638, GU144612; ITS: GU128107). Brazil, Santa Catarina, between GU144624; ITS: GU128126). New Zealand, origin unknown, cultivated Matos Costa and Caçador, cult. University of Michigan Botanical Gardens, material obtained from Tropica Aquarium Plants, Denmark, Troels 040A , Affolter 102 (MICH, GA), DNA accession no. 2660 (cpDNA GU144639, Bone 108 (ILL), DNA accession no. 3178 (cpDNA: GU144651, GU144625; GU144613; ITS: EF177712). Brazil, origin unknown, Casselmann s. n. , 1984, ITS: GU128127). L. occidentalis J. M. Coult. & Rose: U. S. A., California, DNA and leaf material supplied by G. Petersen, Petersen GPL3 (C, ILL), Del Norte Co., S side of the Klamath River, 28 August 1974, Mastroqiuseppe DNA accession no. 2719 (cpDNA: EF185224; ITS: EF177711). South 157 (TEX), DNA accession no. 2974 (ITS: GU128128). U. S. A., California, America, origin unknown, cultivated material identified as L. novae- Marin Co., Bodega Head, July 10, 2007, Fiedler et al. s. n. (SFSU) (ITS: zelandiae obtained from Florida Aquatic Nurseries, Florida, Bone s. n. (ILL), HQ647236). U. S. A., Oregon, Douglas Co., East Gardiner, N of Blacks DNA accession no. 2989 (cpDNA GU144640, GU144614; ITS: GU128108). Island, NE of RR bridge at mouth of Smith River, alluvial sand, gravel and South America, origin unknown, cultivated material obtained from muck, Hill & Dutton 32982 (ILLS 203634), DNA accession no. 1999 (cpDNA: Tropica Aquarium Plants, Denmark, Troels 040 , Bone 107 (ILL), DNA acces- EF185228; ITS: AY360242). U. S. A., Oregon, Douglas Co., Tahkenitch Lake, sion no. 3179 (cpDNA: GU144641, GU144615; ITS: GU128109). L. bris- along U.S. Hwy. 101, 2 November 1995, Halse 5010 (ASU), DNA accession banica A. R. Bean: Australia, Queensland, Brisbane River, SW of Brisbane, no. 2978 (ITS: GU128129). U. S. A., Oregon, Lincoln Co., Kernville, in mud 19 October 1993, Bean 6778 (MO 05061944), DNA accession no. 2886 flat off Pacific Ocean, 22 August 1975, Seigler 9820 (ILL), DNA accession no. (cpDNA: GU144607; ITS: HQ822270). L. carolinensis J. M. Coult. & Rose: 2164 (cpDNA: GU144652, GU144626; ITS: GU128130). U. S. A., Washington, Argentina, Corrientes, Dep. Mburucuyá, Estancia Santa María, cult. Klickitat County, Lyle, on a shoal by the mouth of the Klickitat River, 14 University of Michigan Botanical Garden, Affolter 114 (MICH, GA), DNA September 1992, Halse 4555 (TEX), DNA accession no. 2973 (cpDNA: accession no. 2663 (ITS: EF177713). Portugal, Coimbra, cult. University of GU144653, GU144627; ITS: GU128131). U. S. A., Washington, Mason Co., 2011] BONE ET AL.: PHYLOGENY AND BIOGEOGRAPHY OF LILAEOPSIS (APIACEAE) 805

Mason Lake, 25 July 2008, Fiedler et al. s. n. (SFSU) (ITS: HQ647244). L. Cienega Preserve, growing in ditch in marsh, Titus s. n. (ARIZ 356998), polyantha (Gand.) H. Eichler: Australia, New South Wales, Bombala, Oct. DNA accession no. 2951 (ITS: GU128140). U. S. A., Arizona, Santa Cruz 1998, DNA supplied by G. Petersen, Petersen GPL26 (C), DNA accession Co., San Rafael State Park, along the Santa Cruz River, 2001, McLaughlin & no. 2734 (cpDNA: GU144654, GU144628; ITS: GU128132). L. ruthiana Lewis 9434 (ARIZ 359233), DNA accession no. 2952 (cpDNA: GU144661, Affolter: New Zealand, Canterbury, Lake Lyndon, cultivated, Affolter 176 GU144635). Mexico, Sonora, Los Fresnos Cienega, 32 miles N of Cananea, (MICH), DNA accession no. 2675 (cpDNA: GU144655, GU144629; ITS: 23 June 1990, Warren, Anderson & Saucedo s. n. (ARIZ 292307), DNA acces- GU128133). L. schaffneriana (Schltdl.) J. M. Coult. & Rose subsp. recurva sion no. 2947 (cpDNA: GU144610; ITS: EF177716). L. schaffneriana subsp. (A. W. Hill) Affolter: U. S. A., Arizona, Cochise Co., Huachuca Mts., schaffneriana : Mexico, , 1988, Laferriere 2208 (ARIZ 305701), Coronado Natl. Forest, Scotia Canyon, in stream, 1993, McLaughlin & DNA accession no. 2949 (ITS: GU128141). Mexico, Tláhuac District, 1 Bowers 6378 (ARIZ), DNA accession no. 2943 (ITS: GU128134). U. S. A., September 1989, Jimenez-Osorino s. n. (TEX), DNA accession no. 2977 Arizona, Cochise Co., Huachuca Mts., Sycamore Canyon above Sycamore (cpDNA: GU144662, GU144636; ITS: GU128142). Spring, SW slope of Lone Mt., Elev. 5650′, 1993, Fishbein & Adondakis 1315 OUTGROUPS: Atrema americanum DC.: U. S. A., Texas, Williamson (ARIZ 306551), DNA accession no. 2944 (cpDNA: GU144656, GU144630). Co., 4 miles S of Jarrell on I-35, 18 May 1988, Nesom & Grimes 6415 (MO U. S. A., Arizona, Cochise Co., Leslie Canyon National Wildlife Refuge, 3691937), DNA accession no. 1467 (cpDNA: EF185207; ITS: AY360232). cult. in greenhouse, 11 August 2005, Bone 101 (ILL), DNA accession no. Cynosciadium digitatum DC.: U. S. A., Illinois, Jackson Co., Shawnee 2982 (cpDNA: GU144657, GU144631; ITS: GU128135). U. S. A., Arizona, National Forest, 27 May 1993, Phillippe 21886 (ILLS 183947), DNA acces- Cochise Co., Ft. Huachuca Military Reservation, near Sierra Vista, Garden sion no. 1804 (cpDNA: EF185220; ITS: AY360237). Neogoezia minor Canyon Cienega, 11 August 2005, Bone 102 (ILL), DNA accession no. 2983 Hemsl.: Mexico, , Sierra de San Felipe between Oaxaca and Ixtlán (cpDNA: GU144658, GU144632; ITS: GU128136). U. S. A., Arizona, Cochise de Juárez, 1 August 1963, Molseed 278 (ISU 1060), DNA accession no. 2138 Co., Ft. Huachuca Military Reservation, near Sierra Vista, Garden Canyon, (cpDNA: EF185234; ITS: AY360244). Ptilimnium nuttallii (DC.) Britt.: 11 August 2005, Bone 103 (ILL), DNA accession no. 2986 (cpDNA: U. S. A., Mississippi, Monroe Co., ca. 3 miles W of Aberdeen, 6 June 1996, GU144659, GU144633; ITS: GU128137). U. S. A., Arizona, Cochise Co., Ft. MacDonald 9514 (MO 05082318), DNA accession no. 2405 (ITS: EF177757). Huachuca Military Reservation, near Sierra Vista, Sawmill Springs, 11 U. S. A., Oklahoma, Rogers Co., Claremore, 12 June 1974, Jones 3030 (ILL), August 2005, Bone 104 (ILL), DNA accession no. 2987 (cpDNA: GU144660, DNA accession no. 2165 (cpDNA: EF185259). Trepocarpus aethusae Nutt. GU144634; ITS: GU128138). U. S. A., Arizona, Cochise Co., Coronado ex DC.: U. S. A., Illinois, Alexander Co., Horseshoe Lake Conservation National Memorial, McClure Spring, 12 August 2005, Bone 105 (ILL), DNA Area, 8 July 1996, Basinger 10891 (ILLS 194558), DNA accession no. 1817 accession no. 2988 (ITS: GU128139). U. S. A., Arizona, Pima Co., Bingham (cpDNA: EF185280; ITS: AY360264).